Source: https://openrunners.io/best/

# The cheapest hosted GitHub Actions runners, November 2026

*List price, checked 7 November 2026*

Ranked by published per-minute price for Linux x64 at 2 vCPU. Sengi Basic is lowest at $0.001 a minute ($0.002 at 4 vCPU, $0.004 at 8), then Ubicloud at $0.00125, compiler.dev at $0.002 and Latchkey at $0.0025. GitHub-hosted is $0.006, $0.012 and $0.022. The [benchmark](https://openrunners.io/benchmark/) has every runner and size; [benchmark.json](https://openrunners.io/data/benchmark.json) is the same numbers.


| Runner | 2 vCPU | Free allowance | Billing |
| --- | --- | --- | --- |
| Sengi Basic | $0.001 | 3,000 minutes a month | Per minute, each job rounded up once; boot and teardown unbilled |
| Ubicloud | $0.00125 | $2.50 of credit a month | Per minute |
| compiler.dev | $0.002 | None published | Per minute, from prepaid credit |
| Latchkey | $0.0025 | None; plans from $5 a month | Per minute, rounded up per job |
| Blacksmith | $0.004 | 3,000 minutes a month | Per minute |
| WarpBuild | $0.004 | None published | Per minute; partial-minute rounding not stated |
| GitHub-hosted | $0.006 | Included minutes by plan (private repos) | Per minute, each job rounded up |

Seven of thirteen hosted rows. Namespace and Depot list $0.006 at 2 vCPU; StarSling and Avrea list $0.004. Every row and source link is in the [benchmark](https://openrunners.io/benchmark/).

## What “cheapest hosted” leaves out

A spot machine in your own AWS account can cost less per minute than any hosted runner. That row is in the benchmark. It is your AWS bill: the instance, disk, a public address, and the chance the machine is interrupted. Hetzner bills a started hour, so a 10-minute job pays for the hour. Those are not hosted runners. RunsOn is software you run in your own AWS account, plus a licence, so it is not in this hosted ranking either.

## Price is not speed

A per-minute rate says nothing about how long the job takes. On this site's run of 11 November 2026, the same server test shard at 2 vCPU cost $0.004 on Sengi Basic (3m 33s), $0.00625 on Ubicloud (4m 15s), $0.012 on Sengi Premium (3m 22s) and $0.03 on GitHub-hosted (4m 32s). The cheapest rate was also faster than GitHub's, so the gap per job is wider than the gap per minute. The [ranking](https://openrunners.io/best-github-actions-runners/) has both jobs, the queues, and the raw runs.

For heavier builds, RunsOn's public runner benchmark of 9 October 2026 built a production Docker image at 4 vCPU in 7 minutes 34 seconds on Sengi Premium and 14 minutes 36 seconds on GitHub's 4 vCPU runner. That was a different job and harness, so it is reported here, not mixed into this site's numbers.

## Questions

**What is the cheapest hosted Linux x64 GitHub Actions runner?**
On published list price, Sengi Basic: $0.001 a minute at 2 vCPU, $0.002 at 4 vCPU and $0.004 at 8 vCPU. Ubicloud is next at $0.00125 a minute at 2 vCPU. GitHub-hosted is $0.006, $0.012 and $0.022. Prices were checked on 7 November 2026.

**Is a cheaper per-minute price the same as a cheaper job?**
No. A faster machine can finish in fewer billed minutes, and billing rules change what a short job costs. On 11 November 2026 the same server test shard at 2 vCPU cost $0.004 on Sengi Basic, $0.00625 on Ubicloud and $0.03 on GitHub-hosted.

**Which hosted runners have a free allowance?**
Sengi Basic and Blacksmith each include 3,000 free minutes a month, Avrea 3,000 at 2 vCPU, and Ubicloud $2.50 of credit a month. Depot's Developer plan is $20 a month with 2,000 minutes. GitHub includes minutes by plan for private repositories.

**Which hosted runner is fastest?**
On this site's server test shard at 2 vCPU, Sengi Premium took 3 minutes 22 seconds, Sengi Basic 3 minutes 33 seconds, Ubicloud 4 minutes 15 seconds and GitHub-hosted 4 minutes 32 seconds. Other hosted runners have not been measured yet.

**Can my own AWS or Hetzner machine cost less?**
Yes. A spot instance in your own AWS account can be cheaper per minute than a hosted runner, and it is a different bill: the instance, disk, a public address, and interruptions. Those rows are in the benchmark. They are not hosted runners.

[Open the benchmark](https://openrunners.io/benchmark/) [Switch a job](https://openrunners.io/quickstart/#ci)


openrunners is maintained by engineers who also build Sengi (https://sengi.run/), one of the runners listed.
